Research On The Effectiveness Of The Timing Of Breakout Points、Securities Volume And Volatility Based On Pattern Recognition Technical Analysis

Technical analysis and fundamental analysis are two of the earliest and most distinctive theories in the field of securities investment analysis.Technical analysts believe that prices move along a trend until the game balance between supply and demand forces is broken.The profitability of technical analysis is strongly linked to market efficiency issues in financial markets.The most commonly used tools for technical analysis are charts and data,which utilize historical price data,historical trading volume data,and various behavioral data to speculate on the future development status of the securities market.Technical analysis can be roughly divided into two categories.The first involves constructing strategies based on indicators,with machine learning being widely used in this way.The other involves conducting pattern recognition first,and then constructing a strategy based on the recognized form.The research in this paper is based on the technical analysis of the second type of pattern recognition.Before pattern recognition,the key points of the graphics must be identified,including the time of nodes and the price of key nodes.In the past,traders typically handled this identification process manually during securities trading.However,with the rapid development of computer computing power.the processing of time series data before pattern recognition is now handled by mathematical algorithms.There are roughly two types of identification methods for core points:non-linear curve description(i.e.,non-parametric kernel regression)and approximate identification based on segmentation.Technical analysis based on pattern recognition has been proved in many markets,including the verification of individual stocks and the foreign exchange market in the US market.Empirical papers on the existing ten technical analysis models have also appeared in the Chinese individual stock market in recent years.Among the empirical results of a large number of pattern recognition studies on securities markets in different countries,the similarity lies in that reversal patterns such as headshoulders-top-head-shoulders-bottom perform better in obtaining excess returns,while other patterns have not brought similar excess returns stably and effectively.Based on this background,this article explores the technical analysis of the second type of pattern recognition in the Chinese market in depth through four directions.First,it verifies the broader Chinese market by analyzing the daily data of the Chinese stock index futures market and high-frequency data at the 5-minute level,in addition to individual stocks.Second.more morphological definitions are added,including four new morphologies,and the selection method of the original window width is improved so that the strategy can be dynamically determined according to the estimated length of the morphological shape and the granularity of the data set.Third,the article considers whether the addition of trading volume to the technical analysis of pattern recognition can bring new information gains and changes to profitability for the original strategy through four different strategies.Fourth,the article verifies the breakthrough point strategy for the buying time node of a certain fixed form strategy,and the volatility timing strategy for the selling time node.This paper uses the first type of key point recognition method,i.e.,nonlinear curve description,for the key point recognition before pattern recognition to determine the extreme point and define the basic shape.To test information gain,the paper chooses two statistical methods,KS test and similarity test.To test profitability,different Baseline strategies are used for comparison on the basis of commonly used statistical indicators.Based on the methods outlined in this paper,the following conclusions can be drawn:In the Chinese market,most of the 14 identified new and old patterns in daily data of individual stocks,stock index futures,and high-frequency data provide significant information gain.The head-and-shoulders pattern,in particular,shows a significant reversal effect across various parameters and data,while other patterns may not consistently achieve the expected return effect.The choice of entry time point for different holding time units can greatly affect the profit level shown by different buying strategies.This paper introduces a new definition for 12 breakthrough points of technical forms in the breakthrough point strategy.The new strategies generated based on this definition provide significant incremental information,and the original strategy is significantly enhanced after the breakthrough point is added.The profit effect in the corresponding direction of the reversal effect is drawn.In four attempts to add trading volume to technical analysis based on pattern recognition,the information gain produced by using the noise reduction strategy of trading volume and adding the trading volume factor to the extreme point identification formula of trading volume is significant.However,the ability of trading volume to provide information gain to technical analysis depends not only on how it is applied to the original technical identification,but also on the sensitivity of different technical forms to trading volume.The volatility timing superposition technical analysis trading strategy has significant information gain on both daily data and high-frequency data.The strengthening direction of a specific pattern at different frequencies is generally the same.The strengthening effect brought about by the volatility timing strategy is more significant in daily data than in intraday data.